u14app/deep-research

★ 4,581JavaScriptAudience · generalComplexity · 2/5LicenseSetup · easy

Mindmap

mindmap
  root((deep-research))
    What it does
      Automated research reports
      Multi-step AI loop
      Web search integration
    Inputs
      Questions and topics
      Uploaded PDFs and docs
      Local knowledge base
    Features
      Pause and resume research
      Edit and translate reports
      Knowledge graph view
    Tech Stack
      Next.js
      Tailwind CSS
      Multiple AI providers
    Deployment
      Vercel one-click
      Cloudflare
      Docker

mindmap root((deep-research)) What it does Automated research reports Multi-step AI loop Web search integration Inputs Questions and topics Uploaded PDFs and docs Local knowledge base Features Pause and resume research Edit and translate reports Knowledge graph view Tech Stack Next.js Tailwind CSS Multiple AI providers Deployment Vercel one-click Cloudflare Docker

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Generate a detailed research report on any topic by typing a question and waiting about two minutes.

USE CASE 2

Upload your own PDFs and documents so the AI can answer questions using them as a private knowledge base.

USE CASE 3

Expose a research API endpoint that other apps can call for real-time streaming results via server-sent events.

USE CASE 4

Self-host the entire tool on Vercel or Cloudflare using your own AI provider API keys at no extra cost.

Tech stack

JavaScriptNext.jsTailwind CSSVercelDocker

Getting it running

Difficulty · easy Time to first run · 5min

Needs at least one AI provider API key, a free Gemini key is the quickest path and no backend database is required.

Use, modify, and distribute freely for any purpose including commercial, just keep the MIT copyright notice.

In plain English

Deep Research is a web application that takes a question or topic and produces a detailed research report by querying multiple AI models and web search engines automatically. Instead of the user searching manually and piecing together information, the tool runs a structured research loop in the background and delivers a finished document in roughly two minutes. The application works by combining what the README calls "thinking" and "task" models: one handles reasoning and planning while the other executes individual search steps. It connects to web search services such as Tavily, Brave, and others to fetch current information, then synthesizes the results into a report. Because all data is stored in the browser rather than on a remote server, the user's queries and results stay on their own device. The tool supports a wide range of AI providers. Users can supply their own API keys for Gemini, OpenAI, Anthropic's Claude, Deepseek, Grok, Mistral, Azure OpenAI, Ollama for locally running models, and any service compatible with the OpenAI API format. The quickest way to get started is with a free Gemini API key and a one-click deploy to Vercel or Cloudflare. Beyond basic report generation, there are several additional capabilities. Users can upload PDFs, text files, and Office documents to create a local knowledge base that the AI can reference. Research can be paused and resumed at any stage, or branched in a different direction partway through. Generated reports can be edited in a rich text mode or in Markdown, translated, lengthened or shortened, or converted into a visual knowledge graph. Developers can expose the research functionality as an API endpoint using a server-sent events format, or plug it into other AI tools via a Model Context Protocol server. The project is built with Next.js and Tailwind CSS, deployable via Vercel, Cloudflare, or Docker, and is open source under the MIT license.

Copy-paste prompts

Prompt 1

I want to deploy deep-research to Vercel using my free Gemini API key, walk me through the one-click setup and which environment variables I need to set.

Prompt 2

Using the deep-research server-sent events API endpoint, write a Node.js script that sends a research question and streams the report as it arrives.

Prompt 3

Help me configure deep-research to use a local Ollama model instead of a cloud AI provider, including all environment variables needed.

Prompt 4

I have a folder of PDF reports I want to upload to deep-research as a local knowledge base, show me how to add them and then ask questions against them.

Prompt 5

Show me how to wire deep-research into the Model Context Protocol so I can call it as a research tool from inside Claude or another MCP-compatible assistant.

Open on GitHub → Explain another repo

← u14app on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.